Although there are indications from numerical models that El Niño-Southern Oscillation (ENSO) may be an internal mode of the coupled Pacific ocean-atmosphere system, sensitive to climatic background parameters, it has not yet been possible to find significant changes in ENSO variability between the Little Ice Age and the present. Yet a number of authors have found qualitative indications in anecdotal and proxy records of shorter, century-scale variations in the return-interval statistics for El Niño episodes. To objectively determine what nonstationarities exist, we statistically examine the El Niño occurrences since 1525, compiled by Quinn et al. We have stratified the return intervals both for strong events and for all events according to two null hypotheses. 1) return intervals are stationary over periods of 200–500 years, and 2) the intervals are stationary on a centenary time scale, between epochs of contrasting solar variability. Two-parameter Weibull distributions are fit to subsamples of the data using an optimized bootstrap procedure, and the scale parameters are compared between groups. At the 95% significance level, only the null hypothesis for high/low solar levels and strong El Niño events can be rejected. The corresponding hypothesis for all events rejects at the 90% level, while overall stationarity cannot be rejected at any reasonable level, for either class of events. The significant results are that 1) the El Niño recurrence rate is stationary with respect to long-term climate changes, while 2) return intervals of strong El Niño events are significantly nonstationary at centenary time scales and 3) events of all intensifies exhibit the same nonstationarity but less clearly. There is too little data to reject the possibility that the association with solar epochs is coincidental, however, we have advanced a hypothesis to explain such a connection.